Re: [PATCH v2] arm64: lse: fix LSE atomics with LLVM's integrated assembler

[Will Deacon <will@kernel.org>]

On Tue, Oct 08, 2019 at 02:27:30PM -0700, Sami Tolvanen wrote:
> Unlike gcc, clang considers each inline assembly block to be independent
> and therefore, when using the integrated assembler for inline assembly,
> any preambles that enable features must be repeated in each block.
> 
> This change defines __LSE_PREAMBLE and adds it to each inline assembly
> block that has LSE instructions, which allows them to be compiled also
> with clang's assembler.
> 
> Link: https://github.com/ClangBuiltLinux/linux/issues/671
> Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
> ---
> v2:
>  - Add a preamble to inline assembly blocks that use LSE instead
>    of allowing the compiler to emit LSE instructions everywhere.
> 
> ---
>  arch/arm64/include/asm/atomic_lse.h | 19 +++++++++++++++++++
>  arch/arm64/include/asm/lse.h        |  6 +++---
>  2 files changed, 22 insertions(+), 3 deletions(-)

One thing I've always wanted from binutils is the ability to pass a flag to
the assembler which means that it accepts all of the instructions that it
knows about for a given major architecture (a bit like the '-cpu max' option
to qemu). Even better would be the ability to supply a file at build time
specifying the encodings, so that we could ship that with the kernel and
avoid some of the mess we have in places like sysreg.h were we end up
fighting against the assembler when trying to define new system register
accessors.

The latter suggestion is a bit "pie in the sky", but do you think there is
any scope for the former with clang?

Will